GenAI tools will equip operators with the scalable efficiency gains they need
Operators mastering genAI tools will redefine their operations and business models – while those putting it off will struggle to keep up. That was the clear message at MWC25’s Gen AI Summit: Experimentation to Transformation, which convened leading experts to consider the potential benefits of generative AI technologies to the mobile industry, and the challenges in achieving them.
AI tools in network maintenance are nothing new. Their use is now long established in issue detection, diagnostics and optimisation, boosting efficiency by reducing manual configurations. The present shift sees rapid addition of agentic layers across both new and existing AI capabilities. This development allows customers and staff to guide genAI tools to the right outcome with ease through intuitive prompts. Then, in time, many agentic tools can react autonomously as desired based on the knowledge accrued. AI will increasingly therefore not only analyse data, but also take independent action. There are major gains available here in resolving customer issues, managing network performance, and optimizing workflows. The industry is expecting over 40% of its AI interactions to be conducted autonomously by next year, so clear are these advantages.
How individual operators get there, though, is a bit more complicated. “Every telco board is asking the same questions,” noted Danielle Rios, Acting CEO at Totogi. “How can we deliver shareholder value while spending billions on these new technologies? Telcos are investing heavily in AI, building models and launching initiatives, but transformation is slow. While tech giants launch new features daily, telcos measure progress in months or years.” Why is this?
What’s stopping operators from adopting genAI tools?
There are three major barriers to operators adopting genAI. Firstly, there’s legacy technology debt. The fact is that many telcos operate aging operational systems that can be difficult to integrate with genAI tools. Outdated databases, rigid software architectures, and siloed operations can prevent them from easily drawing on AI’s capabilities. For example, many telcos operate legacy billing and operational support systems which lack the agility needed for AI-driven automation.
There’s also the barrier of data fragmentation. Enterprise data is often scattered across multiple platforms with complicated access requirements. This makes it difficult for genAI tools to access clean, structured information. Without centralised data access, insights from AI systems remain limited, reducing their predictive accuracy and operational impact. Telecom is of course a highly regulated industry in which data privacy concerns are paramount. Telcos often have contracts with defence sectors and other critical industries, making compliance and data sovereignty a top priority which is complex to navigate.
Then there is the reality of cost. Adoption of genAI tools inevitably requires substantial investments in computing power, AI model training, and upskilling of workforces. And while most tech companies operate at a fraction of the overheads of mobile networks, operators spend 80% of their IT budgets just to keep systems running. Some operators therefore understandably hesitate to commit resources to AI projects without clear ROI projections.
How can barriers to adoption be reduced?
Addressing these challenges requires comprehensive digital transformation roadmaps, integrating AI into core enterprise functions, rather than treating it as a set of standalone innovations. Cloud-based AI solutions and AI-as-a-service models can also help to mitigate ROI concerns. Totogi’s BSS Magic platform for instance connects to multiple systems which would otherwise operate separately (such as marketing and order management) through digital twins to plan and execute product launches. What would traditionally be a six-month project can be achieved in minutes, while eliminating huge integrations costs.
According to Anurag Sahay, Head of AI and Data Science at Nagarro, balance must also be struck in selecting model size. Earlier assumptions that larger LLMs are more effective – due to more available data, parameters and compute power – are worth re-examining. “Bigger can actually mean less economical,” noted Sahay. “Smaller models are faster and cheaper – we get there through quantisation, pruning and knowledge distillation.” SLMs will therefore increasingly be the answer to scalability and thereby cost-effectiveness.
What’s on offer for operators investing in genAI tools?
There are three operational areas in which operators are now making most use of genAI tools. Perhaps the most obvious in everyday life is autonomous genAI agents. Agentic genAI tools significantly improve customer engagement by enabling rapidly personalised experiences based on data-driven insights. GenAI chatbots can also handle repetitive queries, automating low-value tasks and reducing call centre workloads. This allows staff to focus on areas that require closer human involvement.
AI-powered recommendations can then enhance sales opportunities through upselling and cross-selling based on relevant user and market data. “Right now your marketing data sits in one stack, and your sales data in another. AI unifies them,” explained Shizana Munir, Solution Engineer at Salesforce. “So when a customer calls, the agent knows which store they visited yesterday and why.” And, more strategically, such tools can help to predict customer churn, suggesting proactive retention strategies and personalised customer journeys.
The days of customers lamenting a ‘robot’ call, then, are receding. Ultimately customers want a quick solution to the issue that brought them to the phone in the first place – and genAI tools are now reducing handling times by 64%. Advances in multilingual capabilities are also expanding these conveniences to a far broader customer base. Agentic genAI assistants not only allow for faster query resolution, though.
Agentic genAI tools go well beyond chatbots
GenAI tools can help operators understand and predict customer satisfaction trends at the same time, through sentiment analysis capabilities. There are more systemic uses for these new technologies in the mobile industry, then. They are now being deployed with remarkable effectiveness in enterprise-level business operations – and not only to improve speed and reduce overheads. Those are of course key advantages here too – for instance with genAI in HR to automate onboarding and training for employees.
GenAI tools are also however proving highly effective at driving down business risk factors. AI-driven billing automation for instance allows highly accurate financial forecasting and fraud detection. AI-based demand forecasting and logistics management are also enhancing supply chain efficiency. And of course network optimisation will remain high on the agenda for most operators, on which AI can now take the reins rather than acting as a mere lens.
Orange’s CTO Laurent Leboucher set out some compelling results from what the operator calls Graph RCA (root cause analysis). This project has allowed Orange to predict battery failures, optimise field service operations, and automatically allocate bandwidth based on real-time demand via dynamic network scaling. “We’ve implemented AI-powered energy optimization, reducing power consumption by switching off unused segments in real time. Graph RCA helps detect network issues by modelling the entire network as a graph. We use AWS-managed services to pinpoint problems instantly.”
Cloud-based AI solutions and interoperability are key to scalability
Cloud integrations will underpin many such projects. Ericsson’s System Comprehension Lab, for example, combines genAI tools into an agentic platform to help engineers grasp complex systems at speed, reduce planning time and become productive more quickly. This collaboration with AWS has enabled developers to save on average one day per week in redundant meetings and emails. A 20% efficiency boost is an extraordinary achievement for a work in progress. This is especially important in an industry expecting only 2.9% CAGR through 2028. Scalable efficiency savings are truly vital.
“These technologies are going to scale quicker than anything that came before,” observed Harry Singh, Chief Digital Officer at BT. “But we can’t let that introduce unmanageable complexity – at BT we’ve reduced our application stack by 40%, and are aiming for an 80% reduction in our tech stack.” Interoperability is essential here – genAI tools must integrate seamlessly across platforms to realise their potential efficiency gains.
Beyond operational efficiency, AI is opening entirely new revenue streams. This is helping redress the historical imbalance in which telcos provided connectivity while tech giants capitalised on digital services. AI is enabling telcos to reclaim lost ground by offering AI-driven sovereign cloud solutions, catering to enterprise customers with localised AI models. AI-powered subscription plans now also allow some operators to bundle genAI digital assistants with data plans, offering businesses intelligent customer service solutions.
According to NVIDIA 85% of operators are now finding ways to drive new revenue through AI, and 75% say they’re doing more with less. “In twelve months, there will be two types of telcos—those explaining why transformation still takes years, and those showing how AI is delivering real value,” warned Danielle Rios. The choice is clear – it’s time for operators to invest in genAI tools, if they are not already. Those who do will find they have chosen the right moment: to overhaul legacy systems, improve customer and employee satisfaction, and take the lead on autonomous network management. Those who do not will find themselves needing to catch up.